Performance evaluation and personalized electric field prediction of the deep H1 coil in the human brain based on simulation and machine learning

Deep transcranial magnetic stimulation (DTMS) has been increasingly used to treat neurological disorders in recent years. However, owing to the complicated configuration of DTMS coils, such as the H1 coil, the electric field induced by it in the personalized human brain is so varied and complex that its transcranial magnetic stimulation performances, especially focusing behavior and depth characteristics, have to be studied and evaluated further before clinical application. Therefore, besides the effects of the excitation frequency of the H1 coils, two types of magnetic shielding blocks (MSBs) with various dimensions were analyzed, and the H1 coil circuit structure with flexible length adjustment and its coil spacing were also investigated in this study. Finally, a machine learning model based on an optimizable tree algorithm was established to rapidly predict the induced electric field in the personalized human brain. Results demonstrated that the half-value depth D_1/2 of the electric field induced by the H1 coil could reach 3.67 cm, which was deeper than that by the figure-of-eight (FOE) coil (<1.6 cm), but its focusing (half-value) volume V_1/2 was 567.94 cm³, larger than that of the FOE coil. After introducing MSBs, reasonably adjusting the coil circuit length and the coil spacing, V_1/2 was reduced to 81.748 cm³, with a slight increase in D_1/2. The proposed machine learning model exhibited a good prediction performance (R² = 0.99, etc.) and only took about 0.014 s to finish predicting the induced electric field in the personalized human brain for rapidly evaluating the H1 coil performance in clinical practices.